Data Description

Description of population

The SEPAGES cohort includes 484 pregnant women. The main analysis is to be done on the continuous variables , that have the compound level in \(µg/L\) with filled-in values below detection limits (< LOD) and standarization for protocol variables. The pooled urine samples are taken during the second and the third trimester of pregnancy, so each woman has two measurements for each of the compounds. There are 477 samples corresponding to the second trimester while there are 454 women who have both the urine samples for the second and third trimester. 384 of the individuals who have the second trimester urine samples have all three thyroid hormone measurements (T4T, T3T and TSH) and do not have known thyroid problems (hypothyroidism, hyperthyroidism or goiter).

Endocrine Disruptors

Thyroid Hormones

For the 437 women who do not have thyroid problems and have the second trimester endocrine disruptor samples, there are 411 samples for Total T4 (TT4) and TSH and, 384 samples for Total T3 (TT3). Only TT4 concentration appears to approach a normal distribution. This is not the case for TT3 or TSH (Appendix).

Thyroid hormones and endocrine disruptors

Confounder Variables

The mothers have a median age of 32.0711823, which is older than the average age of primiparous mothers in France, 30.7 years old (INSEE, 2019). Around 20% of the women had the serum sampling much later than the others. This group will not be used for the main analysis but for the latter sensitivity analysis.

46.0% of the mothers included in the cohort were nulliparous before the currently studied offspring. A similar percentage was primiparous (44.0%) while only 9.7% had already given birth to 2 or more offsprings. 56% of the participants have a postgraduate degree, 27% have an undergraduate degree and 17% have a High school diploma or less.

Protocol Variables

   Min. 1st Qu.  Median    Mean 3rd Qu.    Max.    NA's 
0.02014 0.07083 0.12951 0.35327 0.63889 3.68056       1 

Thyroid Hormones and Confounder Variables

Some of the thyroid hormones have a slight correlation with the confounder variables. The positive correlation between the BMI and the TT3 is the highest (\(r=\) 0.2392242 ). The gestational age at serum sampling has a positive correlation with FT4 (\(r=\) -0.2314729 ) and with FT3 (\(r=\) 0.2392242 ). In addition, the mother’s age has a negative correlation with TT4 (\(r=\) -0.1716068 ).

For the second trimester samples on women without thyroid problems, we studied the significance of univariate regression tests on the four possible confounders for TT3, TT4 and TSH. TT3 has a significant slope on the univariate linear regression on selenium and there is a significant effect of batch on TT3 and TT4, all of which have a p-value under the chosen threshold of 0.2. When this significance tests are done on the entire population, including the women with thyroid problems, iodine concentration also has a significant effect on TT4.

When we study the linear regression of iodine concentration on each of the three thyroid hormones (TT3, TT4 and TSH), the slope has a p-value inferior to the threshold of 0.2 for TT3 and TT4. These results align with Ahad and Ganie. * DOROTHY: : According to Ahad and Ganie, iodine uptake into the follicular cells could be considered the first step in synthesis of thyroid hormones (T3 and T4) (Ahad and Ganie, 2010). The inclusion of iodine in the model depends on which of the three relationship models we consider.

The batch in which the thyroid hormones were measured appears to have a significant effect on the measurement for TT3 (p-value=2.0310^{-4}) and for TT4 (p-value=2.1910^{-9}) but not for TSH (p-value=0.3002).